I think for Java they successfully added support for 80-bit floats?! Anyway, for Julia, I was pretty sure there was a package adding such support, but I can’t track it down. It should be the fastest option for Intel compatible, of options for that much (or more) precision (while not working on e.g. ARM), still slower than Float64, that can be SIMD-enabled. I would really look into other options, because even if I had found that package, it would be platform dependent (or possibly revert to Float64, then presicion not to be relied upon), unlike other better options:
This might be the fastest large-precision package, with more than 80-bit, and cross-platform (would be a bit slower than 80-bit, or possibly not since I guess could be SIMD-enabled?):
https://juliahub.com/docs/MultiFloats/k11Ei/0.4.0/
Two new keywords were added to JDK 1.2 with the beta 4 release:
strictfpandwidefp. These keywords let you specify the IEEE 754-standard strictness that a method or class uses when calculating intermediate results. Currently, these options are not used by the JVM, but you can place them in your code. Thestrictfpkeyword acts like the currentfloatbehavior, whilewidefpis an extended-precision format, which may be faster.
I’m not sure what to read into this “fixed” issue (i.e. “Currently, these options are not used by the JVM” above outdated?):
compiler incorrectly accepts strictfp/widefp constructor
https://bugs.openjdk.java.net/browse/JDK-4153038
I would really consider other options than more precision, such as interval packages, or even posits (valids), if you’re adventurous:
https://news.ycombinator.com/item?id=20388029
https://news.ycombinator.com/item?id=23821030
https://comp.lang.forth.narkive.com/VR4kCtvc/the-performance-cost-of-extended-precision